I am doing a time to event analysis in R. My problem is as follows:

A customer signs a contract at a company for one year and is able to cancel the contract at any time he wants. I am interested in the probability that a customer cancels the contract by himself (so before the automatic cancellation after one year.)

The definition of the 'event' in this case is the cancellation due the client himself. I know that after 365 days (366 for a leap year) the observation is censored. This means that the requirement of noninformative censoring is not met (right?). How can I account for censoring with a fixed end-time in a Cox model?

J.P. Klein and M.L. Moescherger call this type of censoring progressive Type I Censoring in their book Survival Analysis: Techniques for Censored and Truncated Data, however Google provides mostly information about progressive type II (not I) censoring.

  • $\begingroup$ Are you interested in the time to cancellation or only whether cancellation occurs within one year by the customer? $\endgroup$ – Todd D Sep 9 '16 at 22:42
  • $\begingroup$ @Todd I am interested in whether the cancellation occurs due to the client. The client gets a new contract offer after one year and can decide to not accept it. If he doesn't accept, the 'death' is at roughly one year, but it is also possible to cancel the contract earlier. So to answer your question: both (I think) $\endgroup$ – Marcel10 Sep 10 '16 at 3:11
  • $\begingroup$ Can you give a sample of the data that includes some of the 1 year (re)offers, failure conditions, censoring conditions and covariates? $\endgroup$ – Todd D Sep 10 '16 at 4:05
  • $\begingroup$ @Todd I will do tomorrow, I don't have acces to the data from my home computer $\endgroup$ – Marcel10 Sep 11 '16 at 14:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.